CODEWITHSUNDEEP
r
elieli
elieli

Reputation: 41

oddsratio function

How is the oddsratio function in R (from the epitools package, but I have the same problem with odds.ratio from questionr package) calculating the odds ratio? If I do the calculation manually I get a different result. Also, If I input a 2x3 matrix in oddsratio for example, I get a 2x3 vector in return. Which odds ratios are those?

For example, if I create a 2X2 matrix m and calculate the odds ratio manually and compare it to the results of oddsratio:

m = matrix(c(1,6,3,1), ncol = 2, nrow = 2)
odds_rati_manual = ((1/4)/(3/4))/((6/7)/(1/7)) ## = 1/18 = 0.0556
oddsratio(m)$measure = 0.085075

why are the two different?

Also, if I input a matrix with 3 rows:

m = matrix(c(1,6,3,1,4,8), ncol = 2, nrow = 3)
oddsratio(m)$measure

I get an output with three rows. Is that the odds ratio of the odds from the first row to all the other rows?

Upvotes: 1

Views: 642

Answers (1)

G. Grothendieck
G. Grothendieck

Reputation: 270348

oddsratio supports 4 methods.

library(epitools)

args(oddsratio)
## function (x, y = NULL, method = c("midp", "fisher", "wald", "small"), 
##     conf.level = 0.95, rev = c("neither", "rows", "columns", 
##         "both"), correction = FALSE, verbose = FALSE)

Try method = "wald" :

m <- matrix(c(1,6,3,1), ncol = 2, nrow = 2)

((1/4)/(3/4))/((6/7)/(1/7))
## [1] 0.05555556

oddsratio(m, method = "wald")$measure
##           odds ratio with 95% C.I.
## Predictor    estimate     lower    upper
##   Exposed1 1.00000000        NA       NA
##   Exposed2 0.05555556 0.0025053 1.231956 <----------
## Warning message:
## In chisq.test(xx, correct = correction) :
##   Chi-squared approximation may be incorrect

3 rows

Regarding the example with 3 rows each odds ratio is based on the first row and one other row. Compare these:

library(epitools)
m3 <- matrix(c(1,6,3,1,4,8), ncol = 2, nrow = 3)

oddsratio(m3)$measure
##           odds ratio with 95% C.I.
## Predictor  estimate      lower     upper
##   Exposed1 1.000000         NA        NA
##   Exposed2 0.690062 0.01478853  32.24304 <----- compare rows 1 & 2
##   Exposed3 2.449511 0.05171633 115.75523 <----- compare rows 1 & 3

oddsratio(m3[-3, ])$measure
##           odds ratio with 95% C.I.
## Predictor  estimate      lower    upper
##   Exposed1 1.000000         NA       NA
##   Exposed2 0.690062 0.01478853 32.24304 <----- compare rows 1 & 2

oddsratio(m3[-2, ])$measure
##           odds ratio with 95% C.I.
## Predictor  estimate      lower    upper
##   Exposed1 1.000000         NA       NA
##   Exposed2 2.449511 0.05171633 115.7552 <----- compare rows 1 & 3

Upvotes: 3

Related Questions